• FIBROSIS
• CHRONIC HEPATITIS

We read with great interest the letter from Stockdale et al 1 in which they concluded that the gamma-glutamyl transpeptidase to platelet ratio (GPR) and aspartate transaminase (AST)-to-platelet ratio index (APRI) would not be ideal for the diagnosis of liver fibrosis in patients with HIV–HBV coinfection from West Africa. In a relatively small cohort of HIV–HBV coinfected subjects (n=100) from Ghana, the authors reported poor sensitivity (57%–67%) and specificity (63%–79%) of GRP and APRI in diagnosing advanced fibrosis (≥F3) and cirrhosis (F4). These findings are interesting, but should be interpreted with caution with regards to the internal validity of using transient elastography as the reference, the study population itself and the cut-offs generated as a result.

The GPR was originally developed for the assessment of liver fibrosis in HBV mono-infected individuals in a West African setting.2 It has been recently applied to patients predominately treated for HIV–HBV coinfection by a French team, reporting good diagnostic performance to diagnose cirrhosis (sensitivity 77%, specificity 87%).3 In both studies, the accuracy of GPR was assessed using liver histology as a reference.

Liver histology remains the gold standard to validate non-invasive markers of liver fibrosis in clinical research. In contrast, Stockdale et al used transient elastography (Fibroscan) as their reference of choice, principally based on data availability. To the best of our knowledge, the diagnostic performance of Fibroscan has not been evaluated in coinfected patients living in West Africa, and more importantly, they applied cut-offs derived from a majorly European cohort.1 They also did not specify whether patients were fasting when samples and Fibroscan were taken, which could negatively affect appropriate classification.4 Taken together, there exists a potential for misclassification of fibrosis stages, which could have lent to the poor diagnostic performances observed in their study.

In an attempt to evaluate bias due to misclassification error, we used Fibroscan as a reference to assess the performance of three simple non-invasive markers of fibrosis (GPR, APRI and Fib-4), in a large cohort of treatment-naïve HBV mono-infected subjects in The Gambia (n=1026). We should stress that the use of Fibroscan has been well validated in this context and we applied the optimal cut-off values defined in Gambian patients mono-infected with HBV.2 After excluding 135 patients whose data were used to develop and validate the GPR, 721 of 891 had complete data on laboratory parameters and fasting transient elastography. Median age was 35 years old (IQR: 31–43) and 472 (65.5%) were men. Median alanine aminotransferase (ALT), AST, gamma glutamyl-transpeptidase (GGT) values and platelets count were 24 IU/mL (IQR: 19–32), 30 IU/mL (25–35), 26 IU/mL (20–37) and 190 10⁹/L (151–239), respectively. According to optimal Fibroscan cut-offs (7.9, 8.2 and 9.5 kPa to diagnose ≥F2, ≥F3 and F4 METAVIR fibrosis, respectively), 617 (85.6%) were categorised as F0–1, 6 (0.8%) as F2, 32 (4.4%) as F3 and 66 (9.2%) as F4. Diagnostic performance is summarised in table 1. GPR, APRI and Fib-4 had good diagnostic performance but GPR provided significantly higher area under the receiver operating characteristic curve than APRI and Fib-4, implying superiority of GPR in predicting significant liver fibrosis and cirrhosis.

From a biological point of view, the combination of GGT and platelet count may not be a good marker of liver fibrosis in patients with HIV infection from Africa. First, thrombocytopenia is a common manifestation of HIV infection on the continent and can easily overestimate the degree of fibrosis. Second, GGT increases with the use of specific antiretroviral agents, such as non-nucleoside reverse transcriptase inhibitors,3 which are frequently prescribed in West Africa. This issue could have explained the much lower GRP cut-offs in our study compared with those proposed by Stockdale et al, where all patients included were undergoing treatment with this class of antiretrovirals.

In conclusion, we have confirmed herein that GPR predicts significant liver fibrosis and cirrhosis well in a large cohort of HBV mono-infected Gambian patients using locally validated Fibroscan measures as a reference. Care must be taken when interpreting the results by Stockdale et al as fibrosis stages could have been misclassified. Nevertheless, the diagnostic accuracy of GPR needs to be assessed in additional cohorts of HIV–HBV coinfected subjects in Africa, using an appropriate reference standard and among patients undergoing different antiretroviral agents with minimal effects on GGT levels or platelet counts.